Interpretable and Explainable Models for Breast Cancer Recurrence Prediction: A Systematic Review

Breast cancer recurrence remains a major clinical challenge with significant implications for patient survival and quality of life. Although numerous prognostic models have been developed, ranging from traditional statistical approaches to modern machine learning and deep learning techniques, their clinical adoption remains limited. A key barrier is the lack of transparency in how predictions are generated, which restricts clinicians’ ability to trust, interpret, and effectively communicate risk estimates to patients. Interpretability and explainability are therefore essential for bridging the gap between model development and clinical implementation. However, to date, no systematic review has comprehensively examined how these approaches are applied in the context of breast cancer recurrence prediction, nor compared their use across different modelling paradigms. Important questions remain regarding which interpretability techniques are most frequently employed, how model explanations are validated against clinical knowledge, and whether more complex artificial intelligence models provide meaningful advantages over established statistical methods. This systematic review aims to address these gaps by identifying, synthesizing, and critically appraising prognostic models for breast cancer recurrence that incorporate interpretability or explainability. Specifically, it will compare intrinsically interpretable statistical models with machine learning and deep learning models enhanced by explainable artificial intelligence techniques. The review will evaluate trade-offs between predictive performance and interpretability, and will provide evidence-based recommendations for model development, validation, reporting practices, and pathways toward clinical implementation. Ultimately, this work seeks to support the responsible and effective integration of predictive models into oncology practice.